Neighborhood Effects of Foreclosures on Detached Housing Sale Prices in Tokyo

Transcription

1 Neighborhood Effects of Foreclosures on Detached Housing Sale Prices in Tokyo Nobuyoshi Hasegawa more than the number in Recently the number of foreclosures including foreclosed office buildings and store buildings throughout Japan is approximately 50,000 60,000 a year. However, in 2009, only the number of foreclosed housings is more than this standard as above. ABSTRACT On the other hand, on March 19 th, 2010, the Japanese government said that as of January 1, 2010, land prices declined at a record 99.6 percent of the 27,410 survey location throughout the nation, Japan faces declining housing prices and increasing foreclosures in There have been no researches in Japan about neighboring effects of foreclosures on housing prices. This paper attempts to identify whether foreclosures reduce the value of nearby properties, and if so, how much foreclosures in fact drive down neighboring property values. This paper uses dataset on sales and foreclosure filings in Tokyo from 2006 to Regression results suggest that foreclosures in the specific range around housing sales reduce the sale prices, and magnitude of the price discount increases with the number of foreclosed properties. which is the highest ratio of declines since the annual land price survey began in In Tokyo, land prices decreased 6.8 percent on average for residential locations. There are likely to be some effects of increasing number of foreclosures on prices of housing transactions. For example, after the completion of foreclosure proceedings, the property may sit vacant and, signal that the neighborhood is not stable. Properties with distressed loans are likely to sell at a discount. Foreclosure is usually sold at a discount to compensate for the moral risk, amenity deterioration and other negative features associated with foreclosure, as well as a typically shortened marketing period. This paper will examine the effects of foreclosures around housing transactions on the price of Keyword: Foreclosure, Housing sale price, Neighborhood effects, Hedonic analysis the transactions, and consists of five parts. I will explain previous empirical research, foreclosure system in Japan, research hypotheses, data and empirical strategy, and regression results. 1. INTRODUCTION 2. PREVIOUS EMPIRICAL RESEARCH In 2009, the number of foreclosures throughout Japan has enormously increased. This tendency is true of Tokyo. For example, in all 23 wards of Tokyo, the number of foreclosures whose open bids took place in 2009 is more than 1.6 times as the number in (On the other hand, the total number of real estate sales declined from 2008 to 2009.) Foreclosed Real Estate Distribution Association (Fudosan Keibai Ryuutuu Kyokai) discloses that the number of foreclosed housings and apartments throughout Japan, filed between April and September in 2009, amounted to approximately 30,180, which is 46.3% There is no previous research about the effects of foreclosures on neighboring sales in Japan. There are some empirical researches about the regression analysis of housing prices. Hasegawa, Tanishita, Shimizu estimates the effects of landscape regulations on single-family house prices in Setagaya City. As for the analysis of foreclosed properties, Saita analyzes the prices of foreclosed properties in the center regions of Japan since 1993 when the bubble burst. This research concludes that the prices had been

2 declining, and are more responsive to the trends of demand-supply than the published land prices. However, there are no researches that analyze the effects of foreclosures on neighboring real estate sales in Japan. On the other hand, in the United States, there are some empirical researches about the effects of foreclosures on neighboring properties. For example, Immergluck and Smith attempts to estimate the effects of foreclosures of one- to four- family homes on the property values of surrounding one- to fourfamily homes in Chicago and finds that each additional foreclosure within one-eighth of a mile is associated with roughly a one-percent decline in property value. Their study is the first step to use hedonic regression models to estimate the impacts of foreclosure on surrounding property values. Schuetz, Been and Ellen attempts to identify the effects of foreclosures starts on housing prices in the notify the public and the creditors of the time for filing demands, the court marshal (Shikkokan) to review the condition of the property, and a court appraiser to value the property. This usually takes approximately six months. The court determines the minimum price after reviewing the reports from the marshal and the appraiser. Japanese law provides for four types of sale, including a bid (designated date), a period bid, auction, and special sale. The court determines the method of sale, date, time, and place of the auction. The court marshal conducts the auction. Once the highest bid is made, the court will conduct a hearing (baikyaku kettei kijitu) (one week) to decide whether to accept such bid. If accepted, the title of the property passes from the borrower to the purchaser in exchange for the bid price. The proceeds from the sale are then distributed to the guarantor. The entire foreclosure process takes about eight months. surrounding neighborhood in New York City. The results suggest that above some threshold, proximity to properties in foreclosure is associated with lower sale prices. The magnitude of the price discount increases with the number of properties in foreclosure, but not in a linear relationship. 4. RESEARCH HYPOTHESES Properties with distressed loans are likely to sell at a discount both at pre-foreclosure sales 3. FORECLOSURE SYSTEM and at foreclosure auctions thus affecting the price of comparables used to estimate neighboring property values. In Japan, there are two methods for foreclosing on a property: judicial and non-judicial. In judicial process, filing a petition to foreclose takes up to one month and must be placed with the Court of Execution that has jurisdiction. In approximately two to three weeks, an Order of Seizure will be granted. The sale of the property takes place through a court auction. The Court of Execution will authorize the sale of a property through auction procedures (keibai kaishi kettei). Upon authorization, an enforceable In addition, more foreclosed properties add to the supply of local available units, and therefore they depress the values of housing prices in the local area. Moreover, property owners who receive foreclosure notices may be less likely than other homeowners to maintain their properties because they have less incentive to maintain property they might lose, which leads to less valuable properties. instrument is provided to the borrower and the owner on the property. Such instrument is recorded at the registrar. Seizure becomes effective at this point or upon delivery of the seizure notice. The court will

3 5. DATA AND EMPIRICAL STRATEGY property values, they do not completely eliminate it. LP i is a vector of variables indicating the number of LP filings within a given time and distance The empirical analysis attempts to identify the effects of foreclosures on neighboring property values by using a variation of hedonic regression analysis, controlling for property and neighborhood characteristics. The form of the regression is shown below. LPRICE i,j,k = β 0 + β 1 LP i + β 2 PropChars i +β 3 LanduseChars j +β 4 WardChars k +β 5 YearChars l. LPRICE i is the log per square meter sales price of property i. PropChars i is a vector of characteristics describing property i, including land area, age of building, breadth of front road, time to nearest station, time to Tokyo Station. interval of property i. This paper creates a number of measures of proximity to foreclosures. The distance intervals of interest are meter, meter, and meter. 1 The foreclosure process may last up to 8 months in Japan. This paper assumes that the first 8 months after filings are the period that might affect the value of the property near the foreclosed properties. This paper identifies the number of filings within each of those distance bands during the 8 months prior to the sale, and also identifies the number of filings within each of these distance bands more than 8 months prior to the sale, to test whether impacts last longer. LanduseChars j is a vector of characteristics describing land use j around property i, including floor area ratio. TimeDummy l is a time dummy describing year l when property i is transacted. To control for trends in home prices, and to capture how the spillover effects of foreclosures on neighborhood property values may vary over the housing cycles, this paper utilizes data in different years, and includes time dummies. WardDummy k is a ward dummy describing ward k where property i is located. To control for characteristics at the community level, this paper considers ward dummies. The inclusion of ward dummies helps partially to control for the neighborhood differences, but in addition, this paper includes a set of variables indicating the number of foreclosure starts filed after the sale as will discuss below. These future foreclosures would not yet be affecting the value of nearby properties, but they help to proxy for differences in unobserved characteristics between those micro-neighborhoods where foreclosures tend to occur and those where they do not. While our inclusion of future foreclosures and ward dummies should help to reduce the likelihood that our coefficients reflect the propensity of neighborhoods with lower The data set used to estimate the hedonic models contains sales prices, characteristics, and location information for 13,863 detached houses, in 23 wards of Tokyo, that sold between 2006 and All of the housing data were obtained from Athome Inc., a private company which deals with information of real estate transactions. Table 1 shows summary information for these data. The average home has approximately 93.9 square meter, is on a lot of approximately 83.5 square meter, is 5.5years old and sold for just over 539 thousand yen per square meter. The average time to the nearest station is 11.5 minutes, and the average time from the nearest station to Tokyo Station is 30.5 minutes. The average of front road width is 5.1 m, of floor area ratio is 179.8%. To identify whether each sale is likely to be affected by foreclosure proceeding, this research draws on data of foreclosure between 2006 and 2009 in 23 wards of Tokyo, provided by Estate Times, Inc, a private company which collected these data from the district court of Tokyo. The total data size of foreclosed properties is 10,847 for four years. property values to have high rates of foreclosure, rather than the effects foreclosures have on surrounding 1 In Tokyo, one block is approximately 250meter.

4 n= Table 1: Variable descriptive statistics Variable Mean Std. Dev. Min Max Dependent variable Price(yen)/square meter 538, ,290 50,015 7,690,000 Counts of LPs(*) LPs 0-8mos, 0-125m LPs 8+mos, 0-125m LPs post-sale, 0-125m LPs 0-8mos, m LPs 8+mos, m LPs post-sale, m LPs 0-8mos, m LPs 8+mos, m LPs post-sale, m Dummy indicators of LPs Any LPs, 0-8mos, 0-125m Any LPs, 8+mos, 0-125m Any LPs, post-sale, 0-125m Any LPs, 0-8mos, m Any LPs, 8+mos, m Any LPs, post-sale, m Any LPs, 0-8mos, m Any LPs, 8+mos, m Any LPs, post-sale, m Hedonic characteristics Land area (square meter) Building area (square meter) Building age(years) Time to nearest station(min) Time to Tokyo Station(min) Bus dummies = 1 if bus used to the nearest station Front road widths(meter) Floor area ratio *"LPs, 0-8mos, 0-125m" means that "number of LPs within 125meter of sale, 0-8months before sale" As shown in Table 2, the average number of LPs near each sale varies across the time-distance intervals of interest. Only about 11.4 percent of the sales in our dataset are within 125 meter of one or more LPs in the 8 months prior to the sale, and only 0.6 percent of the sales had more than three LPs within that narrowest time-distance interval. By contrast, approximately two-thirds of our sales had at least one LP within meter in the 8 months prior to the sale, and roughly 32 percent of sales had 6 or more LPs in the more than 8 months prior to the sale. In some former specifications this paper uses a simple count of the number of LPs in each time-distance interval, but because this paper does not expect that impacts are linear, in the latter specifications, this paper uses one or more dummy variables indicating the number of LPs in the interval (i.e. 1-5 LPs, 6 or more). Table 2: Number of Sales with Given Number of LPS by Distance/Time Categories LP 0-8months LP>8months LP post sale before sale before sale Number % Number % Number % 0-125m 0LPs LPs LPs m 0LPs

5 1-2LPs LPs LPs m 0LPs LPs LPs LPs LPs Toshima Kita Arakawa Itabashi Nerima Adachi Katsushika Edogawa Setagaya Total As shown in Table 3, annual foreclosure starts during our study period ranged between 2,112 Figure 1: LPs around a sale and 3,626. Figure 1 presents a stylized illustration of the typical sale in our database that is near to a property entering foreclosure. Table 3: Number of LPs by wards and year Total Chiyoda Chuo Minato Shinjuku Bunkyo Taito Sumida Koutou Shinagawa Meguro Ota Shibuya Nakano Suginami This paper wants to control for baseline differences between prices of properties in neighborhood that are vulnerable to foreclosure and properties in neighborhoods that are not. However, of the 13,863 sales in the dataset, only 81 were not within 500 meter of at least one property entering

6 foreclosure between 2006 and 2010, which means that this study has very few sales in the occurrence of future foreclosure starts is correlated with upward housing price. micro-neighborhoods that were completely unaffected by foreclosures. (This is not such an issue within 125 and 250 meter of a sale. There are many sales, 7,428, that do not experience any foreclosure activity within 125 meter, and a reasonable number, 1,731, that do not experience any foreclosure activity within 250 meter.) Thus, this paper will not simply control for baseline differences in price between sales that are within 250 to 500 meter of houses that will enter foreclosures in the future and the few sales that are in neighborhoods which will have no foreclosures during our study period, and instead this paper controls for baseline price differences between sales that are within neighborhoods that will suffer substantial numbers of foreclosures (generally 6 or more) later in the study period and those that will not. Generally, the shorter the distance between a sale and foreclosures is, the greater the effects of foreclosures are. In fact, in New York, according to Schuetz, the coefficient of the feet distance is larger than the one of the feet. However, in Japan, while the coefficient of the m distance is significant, the coefficients of the 0-125m distance and m distance are not significant, which shows that there are no correlations between housing sale prices and the number of foreclosed properties around the houses in the 0-125m and m range. The results could be drawn by the facts that in Japan there are no concentrated foreclosures like New York City, and thus there are no enough foreclosed properties that show the correlation between the numbers of foreclosed properties and the price of housing sales within 125m of sales and in the range of 125 to 250m of sales. For example, as this paper explained above, only about 11.4 percent of the sales in 6. REGRESSION RESULTS our dataset are within 125m of one or more LPs in the 8 months prior to the sale, and only 0.6 percent of the sales had more than three LPs within the interval of 125m in the prior 8 month, and about 26.3 percent Column 1 of Table 4 presents the results of the effects of the number of LPs that occurred in the 8 months prior to the sale within three distance intervals. The coefficients of the 0-125m distance and m distance are not significant. In column 2, this paper adds counts of the number of LPs filed more than 8 months prior to the sale in each distance. Only the coefficient on the number of LPs in the prior 8 months to the sale in the m window is significant and negative. As noted above, these coefficients estimates might be biased, picking up the effects of underlying neighborhood conditions that of the sales in our dataset are within 125 to 250m of one or more LPs in the 8 months prior to the sale, and 3.3 percent of the sales had more than three LPs within the area. On the other hand, in New York City, percent within feet of one or more LPs in the prior 18 months, and 7.0 percent of three or more LPs, 50.7 percent within feet of one or more LPs, and 21.1 percent of three or more LPs. On the contrary, when we analyze the range of 250 to 500m, the number of foreclosures around sales could be said to be enough to show the correlation. are associated with foreclosures. Therefore, in the final column of Table 4, this paper adds counts of LPs that will occur at some point after the sale but within this study period, as an indicator of whether pre-existing neighborhood conditions that may increase the likelihood of foreclosures could be affecting current property values. The coefficient on m post-sale is significant and positive, suggesting that Table 4: Estimated linear impacts of LPs on nearby sales prices Dependent variable Log(price/building area) Variable (1) (2) (3)

7 LPs, 0-8mos, 0-125m (0.001) LPs, 8+mos, 0-125m (0.001) LPs, post-sale, 0-125m (0.001) LPs, 0-8mos, m (0.001) LPs, 8+mos, m (0.000) LPs, post-sale, m (0.000) LPs, 0-8mos, m *** (0.000) ** (0.000) *** (0.000) LPs, 8+mos, m (0.000) LPs, post-sale, m 0.014** (0.000) Observations R-squared Robust standard errors in parentheses ***<0.01, **<0.05, *<0.1 All models include a variety of property, ward and year characteristic. In addition, this paper tests for the quantitative effects of foreclosures on housing prices by It shows the magnitude of the effects of foreclosed properties on the prices of houses nearby, and it also suggests that the more the number of foreclosures around sale is, the greater the effect the foreclosures impose on housing prices is. Table 5: Estimated non-linear impacts of LPs on nearby sales prices Dependent variable Log(price/building area) Variable (1) (2) (3) Any LPs, 0-8mos, m *** (0.001) Any LPs, post-sale, m 0.022** (0.002) 1-5LPs, 0-8mos, m ** ** (0.001) (0.001) 6+LPs, 0-8mos, m *** *** (0.003) (0.003) 3+LPs, post-sale, m 0.024*** (0.002) 1-5LPs, post-sale, m 0.026** (0.002) 6+LPs, post-sale, m 0.033*** (0.003) Observations R-squared Robust standard errors in parentheses ***<0.01, **<0.05, *<0.1 All models include a variety of property, ward and year characteristic. using dummy variables to indicate different numbers of LPs in the prior 8 months. This paper divides the range of LPs into 1-5 LPs and 6 or more LPs in the 8 month prior to sale. The coefficient on 1-5 LPs in Table 6: Coefficient on hedonic variables the 8 month prior to sale is , and the coefficient on 6 or more LPs in the 8 month prior to sale is -

8 Dependent variable Log (price/building area) Property Characteristics Land area (square meter) (log) *** Building age(years) *** (0.000) Time to the nearest station *** (min)(log) (0.003) Time to Tokyo Station(min)(log) *** (0.007) Bus dummies *** (0.002) Front road widths(meter)(log) 0.037*** Land Use Characteristics Floor area ratio(log) *** Ward Characteristics Chiyoda (0.054) Chuo (0.027) Minato 0.059*** (0.015) Shinjuku ** Bunkyo (0.007) Taito *** (0.010) Sumida *** Koutou *** (0.007) Shinagawa *** Meguro 0.038*** Ota *** Shibuya 0.037*** (0.008) Nakano *** (0.004) Suginami *** Toshima *** Kita *** Arakawa *** Itabashi *** Nerima *** Adachi *** Katsushika *** Edogawa *** Year Characteristics TD(2006) 0.166*** TD(2007) 0.236*** TD(2008) 0.206*** TD(2009) 0.065***

9 Robust standard errors in parentheses ***<0.01, **<0.05, *<0.1 REFERENCE Coefficient taken from Table 6, Model (3). The hedonic coefficients do not change substantially across the models shown in Table 5 and 6. Hasegawa, K., Tanishita, M, and Shimizu, C. (2009) Impact of Analysis of Landscape Regulations on Single-Family House prices: Hedonic Analysis in Setagaya City Planning Administration 32 (2), CONCLUSION This regression results provide us some evidence that the prices of housing sales are reduced by the foreclosed properties around the sales. The correlation between the declines of sale prices and the number of foreclosed properties around the sales could be observed only in the specific range of the sales, m of the sales. The value of the coefficient is approximately 1.7 percent, and the more the number of foreclosures around sale is, the greater the effect the foreclosures impose on housing prices is. In order to make use of this research for economic analysis such as predictions of movements of housing sale prices, we need to refine more on the distances between sales and foreclosed properties, the numbers of foreclosed properties around sales, and the characteristics of wards. Saita, Y. (2003) Land Prices in the Tokyo Metropolitan Area: A Hedonic Analysis of Judicial Auction Prices Bank of Japan Working Paper Series No.03- E-4. Leonard, T. and Murdoch, J.C. (2009) The neighborhood effects of foreclosure Journal of Geographic Systems vol. 11, Lin, Z., E. Rosenblatt, and V. W. Yao (2008) Spillover Effects of Foreclosures on Neighborhood Property Values. Journal of Real Estate Finance and Economics 38, Immergluck, D. and G. Smith (2006) The External Costs of Foreclosures: The Impact of Single-Family Mortgage Foreclosures on Property Values. Housing Policy Debate 17(1), Schuetz, J., Vicki Been and Ellen, I.G. (2008) Neighborhood Effects of Concentrated Mortgage Foreclosures Journal of Housing Economics 17,